Figure, a leading name in the field of robotics, has unveiled its latest innovation, a machine learning model named Helix. Brett Adcock, the founder and CEO of Figure, introduced Helix as a "generalist" Vision-Language-Action (VLA) model. Helix stands out by combining visual data with language prompts to control humanoid robots in real time. This groundbreaking model is designed to manage two robots simultaneously, enabling one to assist the other in performing an array of household tasks. The announcement aims to set a new benchmark in the sphere of domestic robotics.
The development of Helix marks a significant shift towards prioritizing home environments in robotics. Figure's approach underscores the importance of making homes a focal point, as opposed to merely concentrating on workplace applications. The company has been showcasing the capabilities of its 02 humanoid robot within various home settings, including kitchens, living rooms, and bathrooms. These environments pose significant challenges because they vary dramatically from one household to another, making it difficult for robots to adapt effectively.
Presently, teaching robots a new behavior demands considerable human input. This process either involves hours of manual programming by experts with PhDs or thousands of demonstration hours. Despite these challenges, Figure remains committed to overcoming these obstacles.
"Hand the bag of cookies to the robot on your right." – Figure
"Receive the bag of cookies from the robot on your left and place it in the open drawer." – Figure
These examples illustrate the potential capabilities that could redefine interactions within homes. Figure emphasizes that for robots to be genuinely useful domestically, they must generate intelligent new behaviors on-demand, especially when dealing with unfamiliar objects.
"For robots to be useful in households, they will need to be capable of generating intelligent new behaviors on-demand, especially for objects they’ve never seen before." – Figure
The current landscape of humanoid robotics is dominated by models like Google DeepMind's RT-2. This model leverages a combination of video and large language models (LLMs) to train robots. However, Figure is carving its own niche by focusing on home applications at an early stage. Despite the complexities involved in adapting to variable home environments and high costs, Figure is determined to make home robotics a reality.
The unveiling of Helix also serves as a strategic move to attract more talent. Figure is actively recruiting engineers to expand its team and bring additional expertise into the project. By showcasing Helix's potential and its current applications with the 02 humanoid robot, Figure aims to draw interest from skilled professionals eager to contribute to this cutting-edge field.
While household robots have not been a priority for most companies due to high costs and adaptability issues, Figure sees immense potential in this market. The company envisions a future where household chores are automated through intelligent humanoid robots. However, this vision is still several years away as it involves refining the technology through rigorous testing and development.
Leave a Reply